diff options
Diffstat (limited to 'indra/llrender')
-rw-r--r-- | indra/llrender/llgl.cpp | 95 | ||||
-rw-r--r-- | indra/llrender/llgl.h | 2 | ||||
-rw-r--r-- | indra/llrender/llglheaders.h | 3 | ||||
-rw-r--r-- | indra/llrender/llglslshader.cpp | 33 | ||||
-rw-r--r-- | indra/llrender/llglslshader.h | 1 | ||||
-rw-r--r-- | indra/llrender/llimagegl.cpp | 43 | ||||
-rw-r--r-- | indra/llrender/llimagegl.h | 7 | ||||
-rw-r--r-- | indra/llrender/llrendertarget.cpp | 4 | ||||
-rw-r--r-- | indra/llrender/llshadermgr.cpp | 125 | ||||
-rw-r--r-- | indra/llrender/llshadermgr.h | 2 | ||||
-rw-r--r-- | indra/llrender/llvertexbuffer.cpp | 491 | ||||
-rw-r--r-- | indra/llrender/llvertexbuffer.h | 62 |
12 files changed, 585 insertions, 283 deletions
diff --git a/indra/llrender/llgl.cpp b/indra/llrender/llgl.cpp index 946e602fee..a819ab0b79 100644 --- a/indra/llrender/llgl.cpp +++ b/indra/llrender/llgl.cpp @@ -94,9 +94,15 @@ void APIENTRY gl_debug_callback(GLenum source, llwarns << "Severity: " << std::hex << severity << llendl; llwarns << "Message: " << message << llendl; llwarns << "-----------------------" << llendl; + if (severity == GL_DEBUG_SEVERITY_HIGH_ARB) + { + llerrs << "Halting on GL Error" << llendl; + } } #endif +void parse_glsl_version(S32& major, S32& minor); + void ll_init_fail_log(std::string filename) { gFailLog.open(filename.c_str()); @@ -295,6 +301,7 @@ PFNGLGETACTIVEUNIFORMARBPROC glGetActiveUniformARB = NULL; PFNGLGETUNIFORMFVARBPROC glGetUniformfvARB = NULL; PFNGLGETUNIFORMIVARBPROC glGetUniformivARB = NULL; PFNGLGETSHADERSOURCEARBPROC glGetShaderSourceARB = NULL; +PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer = NULL; #if LL_WINDOWS PFNWGLCREATECONTEXTATTRIBSARBPROC wglCreateContextAttribsARB = NULL; @@ -443,7 +450,8 @@ LLGLManager::LLGLManager() : mDriverVersionMinor(0), mDriverVersionRelease(0), mGLVersion(1.0f), - + mGLSLVersionMajor(0), + mGLSLVersionMinor(0), mVRAM(0), mGLMaxVertexRange(0), mGLMaxIndexRange(0) @@ -554,6 +562,29 @@ bool LLGLManager::initGL() mGLVersion = mDriverVersionMajor + mDriverVersionMinor * .1f; + if (mGLVersion >= 2.f) + { + parse_glsl_version(mGLSLVersionMajor, mGLSLVersionMinor); + +#if LL_DARWIN + //never use GLSL greater than 1.20 on OSX + if (mGLSLVersionMajor > 1 || mGLSLVersionMinor >= 30) + { + mGLSLVersionMajor = 1; + mGLSLVersionMinor = 20; + } +#endif + } + + if (mGLVersion >= 2.1f && LLImageGL::sCompressTextures) + { //use texture compression + glHint(GL_TEXTURE_COMPRESSION_HINT, GL_NICEST); + } + else + { //GL version is < 3.0, always disable texture compression + LLImageGL::sCompressTextures = false; + } + // Trailing space necessary to keep "nVidia Corpor_ati_on" cards // from being recognized as ATI. if (mGLVendor.substr(0,4) == "ATI ") @@ -574,11 +605,8 @@ bool LLGLManager::initGL() #endif // LL_WINDOWS #if (LL_WINDOWS || LL_LINUX) && !LL_MESA_HEADLESS - // release 7277 is a point at which we verify that ATI OpenGL - // drivers get pretty stable with SL, ~Catalyst 8.2, - // for both Win32 and Linux. - if (mDriverVersionRelease < 7277 && - mDriverVersionRelease != 0) // 0 == Undetectable driver version - these get to pretend to be new ATI drivers, though that decision may be revisited. + // count any pre OpenGL 3.0 implementation as an old driver + if (mGLVersion < 3.f) { mATIOldDriver = TRUE; } @@ -717,6 +745,11 @@ bool LLGLManager::initGL() } #endif + if (mIsIntel) + { //never try to use framebuffer objects on intel (crashy) + mHasFramebufferObject = FALSE; + } + if (mHasFramebufferObject) { glGetIntegerv(GL_MAX_SAMPLES, &mMaxSamples); @@ -1300,6 +1333,7 @@ void LLGLManager::initExtensions() glVertexAttrib4uivARB = (PFNGLVERTEXATTRIB4UIVARBPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttrib4uivARB"); glVertexAttrib4usvARB = (PFNGLVERTEXATTRIB4USVARBPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttrib4usvARB"); glVertexAttribPointerARB = (PFNGLVERTEXATTRIBPOINTERARBPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttribPointerARB"); + glVertexAttribIPointer = (PFNGLVERTEXATTRIBIPOINTERPROC) GLH_EXT_GET_PROC_ADDRESS("glVertexAttribIPointer"); glEnableVertexAttribArrayARB = (PFNGLENABLEVERTEXATTRIBARRAYARBPROC) GLH_EXT_GET_PROC_ADDRESS("glEnableVertexAttribArrayARB"); glDisableVertexAttribArrayARB = (PFNGLDISABLEVERTEXATTRIBARRAYARBPROC) GLH_EXT_GET_PROC_ADDRESS("glDisableVertexAttribArrayARB"); glProgramStringARB = (PFNGLPROGRAMSTRINGARBPROC) GLH_EXT_GET_PROC_ADDRESS("glProgramStringARB"); @@ -2098,6 +2132,55 @@ void parse_gl_version( S32* major, S32* minor, S32* release, std::string* vendor } } + +void parse_glsl_version(S32& major, S32& minor) +{ + // GL_SHADING_LANGUAGE_VERSION returns a null-terminated string with the format: + // <major>.<minor>[.<release>] [<vendor specific>] + + const char* version = (const char*) glGetString(GL_SHADING_LANGUAGE_VERSION); + major = 0; + minor = 0; + + if( !version ) + { + return; + } + + std::string ver_copy( version ); + S32 len = (S32)strlen( version ); /* Flawfinder: ignore */ + S32 i = 0; + S32 start; + // Find the major version + start = i; + for( ; i < len; i++ ) + { + if( '.' == version[i] ) + { + break; + } + } + std::string major_str = ver_copy.substr(start,i-start); + LLStringUtil::convertToS32(major_str, major); + + if( '.' == version[i] ) + { + i++; + } + + // Find the minor version + start = i; + for( ; i < len; i++ ) + { + if( ('.' == version[i]) || isspace(version[i]) ) + { + break; + } + } + std::string minor_str = ver_copy.substr(start,i-start); + LLStringUtil::convertToS32(minor_str, minor); +} + LLGLUserClipPlane::LLGLUserClipPlane(const LLPlane& p, const glh::matrix4f& modelview, const glh::matrix4f& projection, bool apply) { mApply = apply; diff --git a/indra/llrender/llgl.h b/indra/llrender/llgl.h index 6a147b8e19..5a33c98708 100644 --- a/indra/llrender/llgl.h +++ b/indra/llrender/llgl.h @@ -138,6 +138,8 @@ public: S32 mDriverVersionMinor; S32 mDriverVersionRelease; F32 mGLVersion; // e.g = 1.4 + S32 mGLSLVersionMajor; + S32 mGLSLVersionMinor; std::string mDriverVersionVendorString; S32 mVRAM; // VRAM in MB diff --git a/indra/llrender/llglheaders.h b/indra/llrender/llglheaders.h index 10aad202e1..d61ec707f0 100644 --- a/indra/llrender/llglheaders.h +++ b/indra/llrender/llglheaders.h @@ -199,6 +199,7 @@ extern PFNGLVERTEXATTRIB4UBVARBPROC glVertexAttrib4ubvARB; extern PFNGLVERTEXATTRIB4UIVARBPROC glVertexAttrib4uivARB; extern PFNGLVERTEXATTRIB4USVARBPROC glVertexAttrib4usvARB; extern PFNGLVERTEXATTRIBPOINTERARBPROC glVertexAttribPointerARB; +extern PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer; extern PFNGLENABLEVERTEXATTRIBARRAYARBPROC glEnableVertexAttribArrayARB; extern PFNGLDISABLEVERTEXATTRIBARRAYARBPROC glDisableVertexAttribArrayARB; extern PFNGLPROGRAMSTRINGARBPROC glProgramStringARB; @@ -460,6 +461,7 @@ extern PFNGLVERTEXATTRIB4UBVARBPROC glVertexAttrib4ubvARB; extern PFNGLVERTEXATTRIB4UIVARBPROC glVertexAttrib4uivARB; extern PFNGLVERTEXATTRIB4USVARBPROC glVertexAttrib4usvARB; extern PFNGLVERTEXATTRIBPOINTERARBPROC glVertexAttribPointerARB; +extern PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer; extern PFNGLENABLEVERTEXATTRIBARRAYARBPROC glEnableVertexAttribArrayARB; extern PFNGLDISABLEVERTEXATTRIBARRAYARBPROC glDisableVertexAttribArrayARB; extern PFNGLPROGRAMSTRINGARBPROC glProgramStringARB; @@ -693,6 +695,7 @@ extern PFNGLVERTEXATTRIB4UBVARBPROC glVertexAttrib4ubvARB; extern PFNGLVERTEXATTRIB4UIVARBPROC glVertexAttrib4uivARB; extern PFNGLVERTEXATTRIB4USVARBPROC glVertexAttrib4usvARB; extern PFNGLVERTEXATTRIBPOINTERARBPROC glVertexAttribPointerARB; +extern PFNGLVERTEXATTRIBIPOINTERPROC glVertexAttribIPointer; extern PFNGLENABLEVERTEXATTRIBARRAYARBPROC glEnableVertexAttribArrayARB; extern PFNGLDISABLEVERTEXATTRIBARRAYARBPROC glDisableVertexAttribArrayARB; extern PFNGLPROGRAMSTRINGARBPROC glProgramStringARB; diff --git a/indra/llrender/llglslshader.cpp b/indra/llrender/llglslshader.cpp index 6b2852670a..4b7e639aed 100644 --- a/indra/llrender/llglslshader.cpp +++ b/indra/llrender/llglslshader.cpp @@ -64,10 +64,23 @@ BOOL shouldChange(const LLVector4& v1, const LLVector4& v2) } LLShaderFeatures::LLShaderFeatures() -: calculatesLighting(false), isShiny(false), isFullbright(false), hasWaterFog(false), -hasTransport(false), hasSkinning(false), hasObjectSkinning(false), hasAtmospherics(false), isSpecular(false), -hasGamma(false), hasLighting(false), isAlphaLighting(false), calculatesAtmospherics(false), mIndexedTextureChannels(0), disableTextureIndex(false), -hasAlphaMask(false) + : atmosphericHelpers(false) + , calculatesLighting(false) + , calculatesAtmospherics(false) + , hasLighting(false) + , isAlphaLighting(false) + , isShiny(false) + , isFullbright(false) + , isSpecular(false) + , hasWaterFog(false) + , hasTransport(false) + , hasSkinning(false) + , hasObjectSkinning(false) + , hasAtmospherics(false) + , hasGamma(false) + , mIndexedTextureChannels(0) + , disableTextureIndex(false) + , hasAlphaMask(false) { } @@ -96,7 +109,12 @@ void LLGLSLShader::unload() glGetAttachedObjectsARB(mProgramObject, 1024, &count, obj); for (GLsizei i = 0; i < count; i++) { - glDeleteObjectARB(obj[i]); +#if !LL_DARWIN + if (glIsProgramARB(obj[i])) +#endif + { + glDeleteObjectARB(obj[i]); + } } glDeleteObjectARB(mProgramObject); @@ -148,8 +166,9 @@ BOOL LLGLSLShader::createShader(vector<string> * attributes, return FALSE; } - if (gGLManager.mGLVersion < 3.1f) - { //attachShaderFeatures may have set the number of indexed texture channels, so set to 1 again + if (gGLManager.mGLSLVersionMajor < 2 && gGLManager.mGLSLVersionMinor < 3) + { //indexed texture rendering requires GLSL 1.3 or later + //attachShaderFeatures may have set the number of indexed texture channels, so set to 1 again mFeatures.mIndexedTextureChannels = llmin(mFeatures.mIndexedTextureChannels, 1); } diff --git a/indra/llrender/llglslshader.h b/indra/llrender/llglslshader.h index 00b4b0dbd4..7873fe3c4e 100644 --- a/indra/llrender/llglslshader.h +++ b/indra/llrender/llglslshader.h @@ -33,6 +33,7 @@ class LLShaderFeatures { public: + bool atmosphericHelpers; bool calculatesLighting; bool calculatesAtmospherics; bool hasLighting; // implies no transport (it's possible to have neither though) diff --git a/indra/llrender/llimagegl.cpp b/indra/llrender/llimagegl.cpp index 78591ddd38..17131c9d8a 100644 --- a/indra/llrender/llimagegl.cpp +++ b/indra/llrender/llimagegl.cpp @@ -57,6 +57,7 @@ BOOL LLImageGL::sGlobalUseAnisotropic = FALSE; F32 LLImageGL::sLastFrameTime = 0.f; BOOL LLImageGL::sAllowReadBackRaw = FALSE ; LLImageGL* LLImageGL::sDefaultGLTexture = NULL ; +bool LLImageGL::sCompressTextures = false; std::set<LLImageGL*> LLImageGL::sImageList; @@ -477,6 +478,8 @@ void LLImageGL::init(BOOL usemipmaps) mDiscardLevelInAtlas = -1 ; mTexelsInAtlas = 0 ; mTexelsInGLTexture = 0 ; + + mAllowCompression = true; mTarget = GL_TEXTURE_2D; mBindTarget = LLTexUnit::TT_TEXTURE; @@ -705,7 +708,7 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) stop_glerror(); } - LLImageGL::setManualImage(mTarget, gl_level, mFormatInternal, w, h, mFormatPrimary, GL_UNSIGNED_BYTE, (GLvoid*)data_in); + LLImageGL::setManualImage(mTarget, gl_level, mFormatInternal, w, h, mFormatPrimary, GL_UNSIGNED_BYTE, (GLvoid*)data_in, mAllowCompression); if (gl_level == 0) { analyzeAlpha(data_in, w, h); @@ -747,7 +750,7 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) LLImageGL::setManualImage(mTarget, 0, mFormatInternal, w, h, mFormatPrimary, mFormatType, - data_in); + data_in, mAllowCompression); analyzeAlpha(data_in, w, h); stop_glerror(); @@ -805,7 +808,7 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) stop_glerror(); } - LLImageGL::setManualImage(mTarget, m, mFormatInternal, w, h, mFormatPrimary, mFormatType, cur_mip_data); + LLImageGL::setManualImage(mTarget, m, mFormatInternal, w, h, mFormatPrimary, mFormatType, cur_mip_data, mAllowCompression); if (m == 0) { analyzeAlpha(data_in, w, h); @@ -863,7 +866,7 @@ void LLImageGL::setImage(const U8* data_in, BOOL data_hasmips) } LLImageGL::setManualImage(mTarget, 0, mFormatInternal, w, h, - mFormatPrimary, mFormatType, (GLvoid *)data_in); + mFormatPrimary, mFormatType, (GLvoid *)data_in, mAllowCompression); analyzeAlpha(data_in, w, h); updatePickMask(w, h, data_in); @@ -1110,7 +1113,7 @@ void LLImageGL::deleteTextures(S32 numTextures, U32 *textures, bool immediate) } // static -void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels) +void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels, bool allow_compression) { bool use_scratch = false; U32* scratch = NULL; @@ -1173,6 +1176,36 @@ void LLImageGL::setManualImage(U32 target, S32 miplevel, S32 intformat, S32 widt } } + if (LLImageGL::sCompressTextures && allow_compression) + { + switch (intformat) + { + case GL_RGB: + case GL_RGB8: + intformat = GL_COMPRESSED_RGB; + break; + case GL_RGBA: + case GL_RGBA8: + intformat = GL_COMPRESSED_RGBA; + break; + case GL_LUMINANCE: + case GL_LUMINANCE8: + intformat = GL_COMPRESSED_LUMINANCE; + break; + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE8_ALPHA8: + intformat = GL_COMPRESSED_LUMINANCE_ALPHA; + break; + case GL_ALPHA: + case GL_ALPHA8: + intformat = GL_COMPRESSED_ALPHA; + break; + default: + llwarns << "Could not compress format: " << std::hex << intformat << llendl; + break; + } + } + stop_glerror(); glTexImage2D(target, miplevel, intformat, width, height, 0, pixformat, pixtype, use_scratch ? scratch : pixels); stop_glerror(); diff --git a/indra/llrender/llimagegl.h b/indra/llrender/llimagegl.h index 2cfb15b0d9..e23005fe29 100644 --- a/indra/llrender/llimagegl.h +++ b/indra/llrender/llimagegl.h @@ -94,12 +94,13 @@ public: void setSize(S32 width, S32 height, S32 ncomponents); void setComponents(S32 ncomponents) { mComponents = (S8)ncomponents ;} + void setAllowCompression(bool allow) { mAllowCompression = allow; } // These 3 functions currently wrap glGenTextures(), glDeleteTextures(), and glTexImage2D() // for tracking purposes and will be deprecated in the future static void generateTextures(S32 numTextures, U32 *textures); static void deleteTextures(S32 numTextures, U32 *textures, bool immediate = false); - static void setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels); + static void setManualImage(U32 target, S32 miplevel, S32 intformat, S32 width, S32 height, U32 pixformat, U32 pixtype, const void *pixels, bool allow_compression = true); BOOL createGLTexture() ; BOOL createGLTexture(S32 discard_level, const LLImageRaw* imageraw, S32 usename = 0, BOOL to_create = TRUE, @@ -209,6 +210,8 @@ private: U32 mTexelsInAtlas ; U32 mTexelsInGLTexture; + bool mAllowCompression; + protected: LLGLenum mTarget; // Normally GL_TEXTURE2D, sometimes something else (ex. cube maps) LLTexUnit::eTextureType mBindTarget; // Normally TT_TEXTURE, sometimes something else (ex. cube maps) @@ -246,7 +249,7 @@ public: static BOOL sGlobalUseAnisotropic; static LLImageGL* sDefaultGLTexture ; static BOOL sAutomatedTest; - + static bool sCompressTextures; //use GL texture compression #if DEBUG_MISS BOOL mMissed; // Missed on last bind? BOOL getMissed() const { return mMissed; }; diff --git a/indra/llrender/llrendertarget.cpp b/indra/llrender/llrendertarget.cpp index ef2a7395da..780f1dc484 100644 --- a/indra/llrender/llrendertarget.cpp +++ b/indra/llrender/llrendertarget.cpp @@ -143,7 +143,7 @@ bool LLRenderTarget::addColorAttachment(U32 color_fmt) { clear_glerror(); - LLImageGL::setManualImage(LLTexUnit::getInternalType(mUsage), 0, color_fmt, mResX, mResY, GL_RGBA, GL_UNSIGNED_BYTE, NULL); + LLImageGL::setManualImage(LLTexUnit::getInternalType(mUsage), 0, color_fmt, mResX, mResY, GL_RGBA, GL_UNSIGNED_BYTE, NULL, false); if (glGetError() != GL_NO_ERROR) { llwarns << "Could not allocate color buffer for render target." << llendl; @@ -223,7 +223,7 @@ bool LLRenderTarget::allocateDepth() U32 internal_type = LLTexUnit::getInternalType(mUsage); stop_glerror(); clear_glerror(); - LLImageGL::setManualImage(internal_type, 0, GL_DEPTH_COMPONENT24, mResX, mResY, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, NULL); + LLImageGL::setManualImage(internal_type, 0, GL_DEPTH_COMPONENT24, mResX, mResY, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, NULL, false); gGL.getTexUnit(0)->setTextureFilteringOption(LLTexUnit::TFO_POINT); } diff --git a/indra/llrender/llshadermgr.cpp b/indra/llrender/llshadermgr.cpp index 1a03aeebb7..7d384450e6 100644 --- a/indra/llrender/llshadermgr.cpp +++ b/indra/llrender/llshadermgr.cpp @@ -94,13 +94,16 @@ BOOL LLShaderMgr::attachShaderFeatures(LLGLSLShader * shader) } } - if (features->calculatesLighting) + if (features->calculatesLighting || features->atmosphericHelpers) { if (!shader->attachObject("windlight/atmosphericsHelpersV.glsl")) { return FALSE; } + } + if (features->calculatesLighting) + { if (features->isSpecular) { if (!shader->attachObject("lighting/lightFuncSpecularV.glsl")) @@ -572,34 +575,46 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade GLcharARB* text[4096]; GLuint count = 0; - F32 version = gGLManager.mGLVersion; - -//hack to never use GLSL > 1.20 on OSX -#if LL_DARWIN - version = llmin(version, 2.9f); -#endif - - if (version < 2.1f) - { - text[count++] = strdup("#version 110\n"); - text[count++] = strdup("#define ATTRIBUTE attribute\n"); - text[count++] = strdup("#define VARYING varying\n"); - } - else if (version < 3.3f) + S32 major_version = gGLManager.mGLSLVersionMajor; + S32 minor_version = gGLManager.mGLSLVersionMinor; + + if (major_version == 1 && minor_version < 30) { - //set version to 1.20 - text[count++] = strdup("#version 120\n"); - text[count++] = strdup("#define FXAA_GLSL_120 1\n"); - text[count++] = strdup("#define FXAA_FAST_PIXEL_OFFSET 0\n"); - text[count++] = strdup("#define ATTRIBUTE attribute\n"); - text[count++] = strdup("#define VARYING varying\n"); + if (minor_version < 10) + { + //should NEVER get here -- if major version is 1 and minor version is less than 10, + // viewer should never attempt to use shaders, continuing will result in undefined behavior + llerrs << "Unsupported GLSL Version." << llendl; + } + + if (minor_version <= 19) + { + text[count++] = strdup("#version 110\n"); + text[count++] = strdup("#define ATTRIBUTE attribute\n"); + text[count++] = strdup("#define VARYING varying\n"); + text[count++] = strdup("#define VARYING_FLAT varying\n"); + } + else if (minor_version <= 29) + { + //set version to 1.20 + text[count++] = strdup("#version 120\n"); + text[count++] = strdup("#define FXAA_GLSL_120 1\n"); + text[count++] = strdup("#define FXAA_FAST_PIXEL_OFFSET 0\n"); + text[count++] = strdup("#define ATTRIBUTE attribute\n"); + text[count++] = strdup("#define VARYING varying\n"); + text[count++] = strdup("#define VARYING_FLAT varying\n"); + } } else { - if (version < 4.f) + if (major_version < 4) { //set version to 1.30 text[count++] = strdup("#version 130\n"); + + //some implementations of GLSL 1.30 require integer precision be explicitly declared + text[count++] = strdup("precision mediump int;\n"); + text[count++] = strdup("precision highp float;\n"); } else { //set version to 400 @@ -615,16 +630,25 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade { //"varying" state is "out" in a vertex program, "in" in a fragment program // ("varying" is deprecated after version 1.20) text[count++] = strdup("#define VARYING out\n"); + text[count++] = strdup("#define VARYING_FLAT flat out\n"); } else { text[count++] = strdup("#define VARYING in\n"); + text[count++] = strdup("#define VARYING_FLAT flat in\n"); } //backwards compatibility with legacy texture lookup syntax + text[count++] = strdup("#define texture2D texture\n"); text[count++] = strdup("#define textureCube texture\n"); text[count++] = strdup("#define texture2DLod textureLod\n"); text[count++] = strdup("#define shadow2D(a,b) vec2(texture(a,b))\n"); + + if (major_version > 1 || minor_version >= 40) + { //GLSL 1.40 replaces texture2DRect et al with texture + text[count++] = strdup("#define texture2DRect texture\n"); + text[count++] = strdup("#define shadow2DRect(a,b) vec2(texture(a,b))\n"); + } } //copy preprocessor definitions into buffer @@ -648,22 +672,24 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade . uniform sampler2D texN; - VARYING float vary_texture_index; + VARYING_FLAT ivec4 vary_texture_index; + + vec4 ret = vec4(1,0,1,1); vec4 diffuseLookup(vec2 texcoord) { - switch (int(vary_texture_index+0.25)) + switch (vary_texture_index.r)) { - case 0: return texture2D(tex0, texcoord); - case 1: return texture2D(tex1, texcoord); - case 2: return texture2D(tex2, texcoord); + case 0: ret = texture2D(tex0, texcoord); break; + case 1: ret = texture2D(tex1, texcoord); break; + case 2: ret = texture2D(tex2, texcoord); break; . . . - case N: return texture2D(texN, texcoord); + case N: return texture2D(texN, texcoord); break; } - return vec4(0,0,0,0); + return ret; } */ @@ -676,7 +702,7 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade if (texture_index_channels > 1) { - text[count++] = strdup("VARYING float vary_texture_index;\n"); + text[count++] = strdup("VARYING_FLAT ivec4 vary_texture_index;\n"); } text[count++] = strdup("vec4 diffuseLookup(vec2 texcoord)\n"); @@ -688,45 +714,28 @@ GLhandleARB LLShaderMgr::loadShaderFile(const std::string& filename, S32 & shade text[count++] = strdup("return texture2D(tex0, texcoord);\n"); text[count++] = strdup("}\n"); } - else if (gGLManager.mGLVersion >= 3.f) - { - text[count++] = strdup("\tswitch (int(vary_texture_index+0.25))\n"); + else if (major_version > 1 || minor_version >= 30) + { //switches are supported in GLSL 1.30 and later + text[count++] = strdup("\tvec4 ret = vec4(1,0,1,1);\n"); + text[count++] = strdup("\tswitch (vary_texture_index.r)\n"); text[count++] = strdup("\t{\n"); //switch body for (S32 i = 0; i < texture_index_channels; ++i) { - std::string case_str = llformat("\t\tcase %d: return texture2D(tex%d, texcoord);\n", i, i); + std::string case_str = llformat("\t\tcase %d: ret = texture2D(tex%d, texcoord); break;\n", i, i); text[count++] = strdup(case_str.c_str()); } text[count++] = strdup("\t}\n"); - text[count++] = strdup("\treturn vec4(1,0,1,1);\n"); + text[count++] = strdup("\treturn ret;\n"); text[count++] = strdup("}\n"); } else - { - //switches aren't supported, make block that looks like: - /* - int ti = int(vary_texture_index+0.25); - if (ti == 0) return texture2D(tex0, texcoord); - if (ti == 1) return texture2D(tex1, texcoord); - . - . - . - if (ti == N) return texture2D(texN, texcoord); - */ - - text[count++] = strdup("int ti = int(vary_texture_index+0.25);\n"); - for (S32 i = 0; i < texture_index_channels; ++i) - { - std::string if_str = llformat("if (ti == %d) return texture2D(tex%d, texcoord);\n", i, i); - text[count++] = strdup(if_str.c_str()); - } - - text[count++] = strdup("\treturn vec4(1,0,1,1);\n"); - text[count++] = strdup("}\n"); - } + { //should never get here. Indexed texture rendering requires GLSL 1.30 or later + // (for passing integers between vertex and fragment shaders) + llerrs << "Indexed texture rendering requires GLSL 1.30 or later." << llendl; + } } //copy file into memory @@ -1067,6 +1076,8 @@ void LLShaderMgr::initAttribsAndUniforms() mReservedUniforms.push_back("magnification"); mReservedUniforms.push_back("max_cof"); mReservedUniforms.push_back("res_scale"); + mReservedUniforms.push_back("dof_width"); + mReservedUniforms.push_back("dof_height"); mReservedUniforms.push_back("depthMap"); mReservedUniforms.push_back("shadowMap0"); diff --git a/indra/llrender/llshadermgr.h b/indra/llrender/llshadermgr.h index 950e6c9c2f..e28bda6de2 100644 --- a/indra/llrender/llshadermgr.h +++ b/indra/llrender/llshadermgr.h @@ -142,6 +142,8 @@ public: DOF_MAGNIFICATION, DOF_MAX_COF, DOF_RES_SCALE, + DOF_WIDTH, + DOF_HEIGHT, DEFERRED_DEPTH, DEFERRED_SHADOW0, diff --git a/indra/llrender/llvertexbuffer.cpp b/indra/llrender/llvertexbuffer.cpp index eb302392bb..b4899209c4 100644 --- a/indra/llrender/llvertexbuffer.cpp +++ b/indra/llrender/llvertexbuffer.cpp @@ -38,6 +38,8 @@ #include "llglslshader.h" #include "llmemory.h" +#define LL_VBO_POOLING 0 + //Next Highest Power Of Two //helper function, returns first number > v that is a power of 2, or v if v is already a power of 2 U32 nhpo2(U32 v) @@ -49,35 +51,69 @@ U32 nhpo2(U32 v) return r; } +//which power of 2 is i? +//assumes i is a power of 2 > 0 +U32 wpo2(U32 i) +{ + llassert(i > 0); + llassert(nhpo2(i) == i); + + U32 r = 0; + + while (i >>= 1) ++r; + + return r; +} + + +const U32 LL_VBO_BLOCK_SIZE = 2048; + +U32 vbo_block_size(U32 size) +{ //what block size will fit size? + U32 mod = size % LL_VBO_BLOCK_SIZE; + return mod == 0 ? size : size + (LL_VBO_BLOCK_SIZE-mod); +} + +U32 vbo_block_index(U32 size) +{ + return vbo_block_size(size)/LL_VBO_BLOCK_SIZE; +} + + //============================================================================ //static -LLVBOPool LLVertexBuffer::sStreamVBOPool; -LLVBOPool LLVertexBuffer::sDynamicVBOPool; -LLVBOPool LLVertexBuffer::sStreamIBOPool; -LLVBOPool LLVertexBuffer::sDynamicIBOPool; +LLVBOPool LLVertexBuffer::sStreamVBOPool(GL_STREAM_DRAW_ARB, GL_ARRAY_BUFFER_ARB); +LLVBOPool LLVertexBuffer::sDynamicVBOPool(GL_DYNAMIC_DRAW_ARB, GL_ARRAY_BUFFER_ARB); +LLVBOPool LLVertexBuffer::sStreamIBOPool(GL_STREAM_DRAW_ARB, GL_ELEMENT_ARRAY_BUFFER_ARB); +LLVBOPool LLVertexBuffer::sDynamicIBOPool(GL_DYNAMIC_DRAW_ARB, GL_ELEMENT_ARRAY_BUFFER_ARB); + U32 LLVBOPool::sBytesPooled = 0; +U32 LLVBOPool::sIndexBytesPooled = 0; +U32 LLVertexBuffer::sAllocatedIndexBytes = 0; +U32 LLVertexBuffer::sIndexCount = 0; -LLPrivateMemoryPool* LLVertexBuffer::sPrivatePoolp = NULL ; +LLPrivateMemoryPool* LLVertexBuffer::sPrivatePoolp = NULL; U32 LLVertexBuffer::sBindCount = 0; U32 LLVertexBuffer::sSetCount = 0; S32 LLVertexBuffer::sCount = 0; S32 LLVertexBuffer::sGLCount = 0; S32 LLVertexBuffer::sMappedCount = 0; -BOOL LLVertexBuffer::sDisableVBOMapping = FALSE ; -BOOL LLVertexBuffer::sEnableVBOs = TRUE; +bool LLVertexBuffer::sDisableVBOMapping = false; +bool LLVertexBuffer::sEnableVBOs = true; U32 LLVertexBuffer::sGLRenderBuffer = 0; U32 LLVertexBuffer::sGLRenderArray = 0; U32 LLVertexBuffer::sGLRenderIndices = 0; U32 LLVertexBuffer::sLastMask = 0; -BOOL LLVertexBuffer::sVBOActive = FALSE; -BOOL LLVertexBuffer::sIBOActive = FALSE; +bool LLVertexBuffer::sVBOActive = false; +bool LLVertexBuffer::sIBOActive = false; U32 LLVertexBuffer::sAllocatedBytes = 0; -BOOL LLVertexBuffer::sMapped = FALSE; -BOOL LLVertexBuffer::sUseStreamDraw = TRUE; -BOOL LLVertexBuffer::sUseVAO = FALSE; -BOOL LLVertexBuffer::sPreferStreamDraw = FALSE; +U32 LLVertexBuffer::sVertexCount = 0; +bool LLVertexBuffer::sMapped = false; +bool LLVertexBuffer::sUseStreamDraw = true; +bool LLVertexBuffer::sUseVAO = false; +bool LLVertexBuffer::sPreferStreamDraw = false; const U32 FENCE_WAIT_TIME_NANOSECONDS = 10000; //1 ms @@ -134,39 +170,35 @@ public: }; -//which power of 2 is i? -//assumes i is a power of 2 > 0 -U32 wpo2(U32 i) -{ - llassert(i > 0); - llassert(nhpo2(i) == i); - - U32 r = 0; - - while (i >>= 1) ++r; - - return r; -} - volatile U8* LLVBOPool::allocate(U32& name, U32 size) { - llassert(nhpo2(size) == size); + llassert(vbo_block_size(size) == size); + + volatile U8* ret = NULL; + +#if LL_VBO_POOLING - U32 i = wpo2(size); + U32 i = vbo_block_index(size); if (mFreeList.size() <= i) { mFreeList.resize(i+1); } - volatile U8* ret = NULL; - if (mFreeList[i].empty()) { //make a new buffer glGenBuffersARB(1, &name); glBindBufferARB(mType, name); - LLVertexBuffer::sAllocatedBytes += size; + + if (mType == GL_ARRAY_BUFFER_ARB) + { + LLVertexBuffer::sAllocatedBytes += size; + } + else + { + LLVertexBuffer::sAllocatedIndexBytes += size; + } if (LLVertexBuffer::sDisableVBOMapping || mUsage != GL_DYNAMIC_DRAW_ARB) { @@ -185,36 +217,91 @@ volatile U8* LLVBOPool::allocate(U32& name, U32 size) name = mFreeList[i].front().mGLName; ret = mFreeList[i].front().mClientData; - sBytesPooled -= size; + if (mType == GL_ARRAY_BUFFER_ARB) + { + sBytesPooled -= size; + } + else + { + sIndexBytesPooled -= size; + } mFreeList[i].pop_front(); } +#else //no pooling + + glGenBuffersARB(1, &name); + glBindBufferARB(mType, name); + + if (mType == GL_ARRAY_BUFFER_ARB) + { + LLVertexBuffer::sAllocatedBytes += size; + } + else + { + LLVertexBuffer::sAllocatedIndexBytes += size; + } + + if (LLVertexBuffer::sDisableVBOMapping || mUsage != GL_DYNAMIC_DRAW_ARB) + { + glBufferDataARB(mType, size, 0, mUsage); + ret = (U8*) ll_aligned_malloc_16(size); + } + else + { //always use a true hint of static draw when allocating non-client-backed buffers + glBufferDataARB(mType, size, 0, GL_STATIC_DRAW_ARB); + } + + glBindBufferARB(mType, 0); + +#endif return ret; } void LLVBOPool::release(U32 name, volatile U8* buffer, U32 size) { - llassert(nhpo2(size) == size); + llassert(vbo_block_size(size) == size); + +#if LL_VBO_POOLING - U32 i = wpo2(size); + U32 i = vbo_block_index(size); llassert(mFreeList.size() > i); Record rec; rec.mGLName = name; rec.mClientData = buffer; - - sBytesPooled += size; - if (!LLVertexBuffer::sDisableVBOMapping && mUsage == GL_DYNAMIC_DRAW_ARB) + if (buffer == NULL) { glDeleteBuffersARB(1, &rec.mGLName); } else { + if (mType == GL_ARRAY_BUFFER_ARB) + { + sBytesPooled += size; + } + else + { + sIndexBytesPooled += size; + } mFreeList[i].push_back(rec); } +#else //no pooling + glDeleteBuffersARB(1, &name); + ll_aligned_free_16((U8*) buffer); + + if (mType == GL_ARRAY_BUFFER_ARB) + { + LLVertexBuffer::sAllocatedBytes -= size; + } + else + { + LLVertexBuffer::sAllocatedIndexBytes -= size; + } +#endif } void LLVBOPool::cleanup() @@ -238,8 +325,16 @@ void LLVBOPool::cleanup() l.pop_front(); - LLVertexBuffer::sAllocatedBytes -= size; - sBytesPooled -= size; + if (mType == GL_ARRAY_BUFFER_ARB) + { + sBytesPooled -= size; + LLVertexBuffer::sAllocatedBytes -= size; + } + else + { + sIndexBytesPooled -= size; + LLVertexBuffer::sAllocatedIndexBytes -= size; + } } size *= 2; @@ -283,7 +378,13 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask) { if (sLastMask != data_mask) { - BOOL error = FALSE; + bool error = false; + + if (gGLManager.mGLSLVersionMajor < 2 && gGLManager.mGLSLVersionMinor < 30) + { + //make sure texture index is disabled + data_mask = data_mask & ~MAP_TEXTURE_INDEX; + } if (LLGLSLShader::sNoFixedFunction) { @@ -344,7 +445,7 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask) { if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Bad client state! " << array[i] << " disabled." << std::endl; } else @@ -364,7 +465,7 @@ void LLVertexBuffer::setupClientArrays(U32 data_mask) { //needs to be disabled, make sure it was (DEBUG TEMPORARY) if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Bad client state! " << array[i] << " enabled." << std::endl; } else @@ -430,7 +531,7 @@ void LLVertexBuffer::drawArrays(U32 mode, const std::vector<LLVector3>& pos, con U32 count = pos.size(); llassert_always(norm.size() >= pos.size()); - llassert_always(count > 0) ; + llassert_always(count > 0); unbind(); @@ -548,7 +649,7 @@ void LLVertexBuffer::validateRange(U32 start, U32 end, U32 count, U32 indices_of void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indices_offset) const { validateRange(start, end, count, indices_offset); - mMappable = FALSE; + mMappable = false; gGL.syncMatrices(); llassert(mNumVerts >= 0); @@ -603,7 +704,7 @@ void LLVertexBuffer::drawRange(U32 mode, U32 start, U32 end, U32 count, U32 indi void LLVertexBuffer::draw(U32 mode, U32 count, U32 indices_offset) const { llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL); - mMappable = FALSE; + mMappable = false; gGL.syncMatrices(); llassert(mNumIndices >= 0); @@ -649,7 +750,7 @@ void LLVertexBuffer::draw(U32 mode, U32 count, U32 indices_offset) const void LLVertexBuffer::drawArrays(U32 mode, U32 first, U32 count) const { llassert(!LLGLSLShader::sNoFixedFunction || LLGLSLShader::sCurBoundShaderPtr != NULL); - mMappable = FALSE; + mMappable = false; gGL.syncMatrices(); llassert(mNumVerts >= 0); @@ -689,23 +790,13 @@ void LLVertexBuffer::drawArrays(U32 mode, U32 first, U32 count) const //static void LLVertexBuffer::initClass(bool use_vbo, bool no_vbo_mapping) { - sEnableVBOs = use_vbo && gGLManager.mHasVertexBufferObject ; - sDisableVBOMapping = sEnableVBOs && no_vbo_mapping ; + sEnableVBOs = use_vbo && gGLManager.mHasVertexBufferObject; + sDisableVBOMapping = sEnableVBOs && no_vbo_mapping; - if(!sPrivatePoolp) + if (!sPrivatePoolp) { - sPrivatePoolp = LLPrivateMemoryPoolManager::getInstance()->newPool(LLPrivateMemoryPool::STATIC) ; + sPrivatePoolp = LLPrivateMemoryPoolManager::getInstance()->newPool(LLPrivateMemoryPool::STATIC); } - - sStreamVBOPool.mType = GL_ARRAY_BUFFER_ARB; - sStreamVBOPool.mUsage= GL_STREAM_DRAW_ARB; - sStreamIBOPool.mType = GL_ELEMENT_ARRAY_BUFFER_ARB; - sStreamIBOPool.mUsage= GL_STREAM_DRAW_ARB; - - sDynamicVBOPool.mType = GL_ARRAY_BUFFER_ARB; - sDynamicVBOPool.mUsage= GL_DYNAMIC_DRAW_ARB; - sDynamicIBOPool.mType = GL_ELEMENT_ARRAY_BUFFER_ARB; - sDynamicIBOPool.mUsage= GL_DYNAMIC_DRAW_ARB; } //static @@ -718,18 +809,18 @@ void LLVertexBuffer::unbind() #endif sGLRenderArray = 0; sGLRenderIndices = 0; - sIBOActive = FALSE; + sIBOActive = false; } if (sVBOActive) { glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0); - sVBOActive = FALSE; + sVBOActive = false; } if (sIBOActive) { glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0); - sIBOActive = FALSE; + sIBOActive = false; } sGLRenderBuffer = 0; @@ -751,73 +842,80 @@ void LLVertexBuffer::cleanupClass() if(sPrivatePoolp) { - LLPrivateMemoryPoolManager::getInstance()->deletePool(sPrivatePoolp) ; - sPrivatePoolp = NULL ; + LLPrivateMemoryPoolManager::getInstance()->deletePool(sPrivatePoolp); + sPrivatePoolp = NULL; } } //---------------------------------------------------------------------------- -LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : - LLRefCount(), - - mNumVerts(0), - mNumIndices(0), - mUsage(usage), - mGLBuffer(0), - mGLArray(0), - mGLIndices(0), - mMappedData(NULL), - mMappedIndexData(NULL), - mVertexLocked(FALSE), - mIndexLocked(FALSE), - mFinal(FALSE), - mEmpty(TRUE), - mFence(NULL) +S32 LLVertexBuffer::determineUsage(S32 usage) { - LLMemType mt2(LLMemType::MTYPE_VERTEX_CONSTRUCTOR); - mFence = NULL; + S32 ret_usage = usage; + if (!sEnableVBOs) { - mUsage = 0 ; + ret_usage = 0; } - - if (mUsage == GL_STREAM_DRAW_ARB && !sUseStreamDraw) + + if (ret_usage == GL_STREAM_DRAW_ARB && !sUseStreamDraw) { - mUsage = 0; + ret_usage = 0; } - if (mUsage == GL_DYNAMIC_DRAW_ARB && sPreferStreamDraw) + if (ret_usage == GL_DYNAMIC_DRAW_ARB && sPreferStreamDraw) { - mUsage = GL_STREAM_DRAW_ARB; + ret_usage = GL_STREAM_DRAW_ARB; } - - if (mUsage == 0 && LLRender::sGLCoreProfile) + + if (ret_usage == 0 && LLRender::sGLCoreProfile) { //MUST use VBOs for all rendering - mUsage = GL_STREAM_DRAW_ARB; + ret_usage = GL_STREAM_DRAW_ARB; } - - if (mUsage && mUsage != GL_STREAM_DRAW_ARB) + + if (ret_usage && ret_usage != GL_STREAM_DRAW_ARB) { //only stream_draw and dynamic_draw are supported when using VBOs, dynamic draw is the default if (sDisableVBOMapping) { //always use stream draw if VBO mapping is disabled - mUsage = GL_STREAM_DRAW_ARB; + ret_usage = GL_STREAM_DRAW_ARB; } else { - mUsage = GL_DYNAMIC_DRAW_ARB; + ret_usage = GL_DYNAMIC_DRAW_ARB; } } + return ret_usage; +} - if (mUsage == GL_DYNAMIC_DRAW_ARB && !sDisableVBOMapping) - { - mMappable = TRUE; - } - else - { - mMappable = FALSE; - } +LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : + LLRefCount(), + + mNumVerts(0), + mNumIndices(0), + mAlignedOffset(0), + mAlignedIndexOffset(0), + mSize(0), + mIndicesSize(0), + mTypeMask(typemask), + mUsage(LLVertexBuffer::determineUsage(usage)), + mGLBuffer(0), + mGLIndices(0), + mGLArray(0), + mMappedData(NULL), + mMappedIndexData(NULL), + mMappedDataUsingVBOs(false), + mMappedIndexDataUsingVBOs(false), + mVertexLocked(false), + mIndexLocked(false), + mFinal(false), + mEmpty(true), + mMappable(false), + mFence(NULL) +{ + LLMemType mt2(LLMemType::MTYPE_VERTEX_CONSTRUCTOR); + + mMappable = (mUsage == GL_DYNAMIC_DRAW_ARB && !sDisableVBOMapping); //zero out offsets for (U32 i = 0; i < TYPE_MAX; i++) @@ -825,12 +923,6 @@ LLVertexBuffer::LLVertexBuffer(U32 typemask, S32 usage) : mOffsets[i] = 0; } - mTypeMask = typemask; - mSize = 0; - mIndicesSize = 0; - mAlignedOffset = 0; - mAlignedIndexOffset = 0; - sCount++; } @@ -902,7 +994,10 @@ LLVertexBuffer::~LLVertexBuffer() mFence = NULL; - llassert_always(!mMappedData && !mMappedIndexData) ; + sVertexCount -= mNumVerts; + sIndexCount -= mNumIndices; + + llassert_always(!mMappedData && !mMappedIndexData); }; void LLVertexBuffer::placeFence() const @@ -933,7 +1028,7 @@ void LLVertexBuffer::waitFence() const void LLVertexBuffer::genBuffer(U32 size) { - mSize = nhpo2(size); + mSize = vbo_block_size(size); if (mUsage == GL_STREAM_DRAW_ARB) { @@ -949,7 +1044,7 @@ void LLVertexBuffer::genBuffer(U32 size) void LLVertexBuffer::genIndices(U32 size) { - mIndicesSize = nhpo2(size); + mIndicesSize = vbo_block_size(size); if (mUsage == GL_STREAM_DRAW_ARB) { @@ -1011,9 +1106,11 @@ void LLVertexBuffer::createGLBuffer(U32 size) return; } - mEmpty = TRUE; + mEmpty = true; - if (useVBOs()) + mMappedDataUsingVBOs = useVBOs(); + + if (mMappedDataUsingVBOs) { genBuffer(size); } @@ -1040,12 +1137,14 @@ void LLVertexBuffer::createGLIndices(U32 size) return; } - mEmpty = TRUE; + mEmpty = true; //pad by 16 bytes for aligned copies size += 16; - if (useVBOs()) + mMappedIndexDataUsingVBOs = useVBOs(); + + if (mMappedIndexDataUsingVBOs) { //pad by another 16 bytes for VBO pointer adjustment size += 16; @@ -1065,15 +1164,15 @@ void LLVertexBuffer::destroyGLBuffer() LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTROY_BUFFER); if (mGLBuffer) { - if (useVBOs()) + if (mMappedDataUsingVBOs) { releaseBuffer(); } else { - FREE_MEM(sPrivatePoolp, (void*) mMappedData) ; + FREE_MEM(sPrivatePoolp, (void*) mMappedData); mMappedData = NULL; - mEmpty = TRUE; + mEmpty = true; } } @@ -1086,15 +1185,15 @@ void LLVertexBuffer::destroyGLIndices() LLMemType mt2(LLMemType::MTYPE_VERTEX_DESTROY_INDICES); if (mGLIndices) { - if (useVBOs()) + if (mMappedIndexDataUsingVBOs) { releaseIndices(); } else { - FREE_MEM(sPrivatePoolp, (void*) mMappedIndexData) ; + FREE_MEM(sPrivatePoolp, (void*) mMappedIndexData); mMappedIndexData = NULL; - mEmpty = TRUE; + mEmpty = true; } } @@ -1121,7 +1220,9 @@ void LLVertexBuffer::updateNumVerts(S32 nverts) createGLBuffer(needed_size); } + sVertexCount -= mNumVerts; mNumVerts = nverts; + sVertexCount += mNumVerts; } void LLVertexBuffer::updateNumIndices(S32 nindices) @@ -1137,7 +1238,9 @@ void LLVertexBuffer::updateNumIndices(S32 nindices) createGLIndices(needed_size); } + sIndexCount -= mNumIndices; mNumIndices = nindices; + sIndexCount += mNumIndices; } void LLVertexBuffer::allocateBuffer(S32 nverts, S32 nindices, bool create) @@ -1199,7 +1302,7 @@ void LLVertexBuffer::setupVertexArray() 1, //TYPE_WEIGHT, 4, //TYPE_WEIGHT4, 4, //TYPE_CLOTHWEIGHT, - 1, //TYPE_TEXTURE_INDEX + 4, //TYPE_TEXTURE_INDEX }; U32 attrib_type[] = @@ -1216,7 +1319,24 @@ void LLVertexBuffer::setupVertexArray() GL_FLOAT, //TYPE_WEIGHT, GL_FLOAT, //TYPE_WEIGHT4, GL_FLOAT, //TYPE_CLOTHWEIGHT, - GL_FLOAT, //TYPE_TEXTURE_INDEX + GL_UNSIGNED_BYTE, //TYPE_TEXTURE_INDEX + }; + + bool attrib_integer[] = + { + false, //TYPE_VERTEX, + false, //TYPE_NORMAL, + false, //TYPE_TEXCOORD0, + false, //TYPE_TEXCOORD1, + false, //TYPE_TEXCOORD2, + false, //TYPE_TEXCOORD3, + false, //TYPE_COLOR, + false, //TYPE_EMISSIVE, + false, //TYPE_BINORMAL, + false, //TYPE_WEIGHT, + false, //TYPE_WEIGHT4, + false, //TYPE_CLOTHWEIGHT, + true, //TYPE_TEXTURE_INDEX }; U32 attrib_normalized[] = @@ -1244,7 +1364,21 @@ void LLVertexBuffer::setupVertexArray() if (mTypeMask & (1 << i)) { glEnableVertexAttribArrayARB(i); - glVertexAttribPointerARB(i, attrib_size[i], attrib_type[i], attrib_normalized[i], sTypeSize[i], (void*) mOffsets[i]); + + if (attrib_integer[i]) + { +#if !LL_DARWIN + //glVertexattribIPointer requires GLSL 1.30 or later + if (gGLManager.mGLSLVersionMajor > 1 || gGLManager.mGLSLVersionMinor >= 30) + { + glVertexAttribIPointer(i, attrib_size[i], attrib_type[i], sTypeSize[i], (void*) mOffsets[i]); + } +#endif + } + else + { + glVertexAttribPointerARB(i, attrib_size[i], attrib_type[i], attrib_normalized[i], sTypeSize[i], (void*) mOffsets[i]); + } } else { @@ -1279,16 +1413,10 @@ void LLVertexBuffer::resizeBuffer(S32 newnverts, S32 newnindices) } } -BOOL LLVertexBuffer::useVBOs() const +bool LLVertexBuffer::useVBOs() const { //it's generally ineffective to use VBO for things that are streaming on apple - - if (!mUsage) - { - return FALSE; - } - - return TRUE; + return (mUsage != 0); } //---------------------------------------------------------------------------- @@ -1368,7 +1496,7 @@ volatile U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, boo if (!mVertexLocked) { LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_VERTICES); - mVertexLocked = TRUE; + mVertexLocked = true; sMappedCount++; stop_glerror(); @@ -1447,17 +1575,17 @@ volatile U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, boo { log_glerror(); - //check the availability of memory - LLMemory::logMemoryInfo(TRUE) ; + //check the availability of memory + LLMemory::logMemoryInfo(true); if(mMappable) { //-------------------- //print out more debug info before crash - llinfos << "vertex buffer size: (num verts : num indices) = " << getNumVerts() << " : " << getNumIndices() << llendl ; - GLint size ; - glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size) ; - llinfos << "GL_ARRAY_BUFFER_ARB size is " << size << llendl ; + llinfos << "vertex buffer size: (num verts : num indices) = " << getNumVerts() << " : " << getNumIndices() << llendl; + GLint size; + glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size); + llinfos << "GL_ARRAY_BUFFER_ARB size is " << size << llendl; //-------------------- GLint buff; @@ -1472,7 +1600,7 @@ volatile U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index, S32 count, boo } else { - llerrs << "memory allocation for vertex data failed." << llendl ; + llerrs << "memory allocation for vertex data failed." << llendl; } } } @@ -1547,7 +1675,7 @@ volatile U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range { LLMemType mt_v(LLMemType::MTYPE_VERTEX_MAP_BUFFER_INDICES); - mIndexLocked = TRUE; + mIndexLocked = true; sMappedCount++; stop_glerror(); @@ -1626,7 +1754,7 @@ volatile U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range if (!mMappedIndexData) { log_glerror(); - LLMemory::logMemoryInfo(TRUE) ; + LLMemory::logMemoryInfo(true); if(mMappable) { @@ -1641,7 +1769,7 @@ volatile U8* LLVertexBuffer::mapIndexBuffer(S32 index, S32 count, bool map_range } else { - llerrs << "memory allocation for Index data failed. " << llendl ; + llerrs << "memory allocation for Index data failed. " << llendl; } } } @@ -1672,10 +1800,10 @@ void LLVertexBuffer::unmapBuffer() LLMemType mt2(LLMemType::MTYPE_VERTEX_UNMAP_BUFFER); if (!useVBOs()) { - return ; //nothing to unmap + return; //nothing to unmap } - bool updated_all = false ; + bool updated_all = false; if (mMappedData && mVertexLocked) { @@ -1742,7 +1870,7 @@ void LLVertexBuffer::unmapBuffer() mMappedData = NULL; } - mVertexLocked = FALSE ; + mVertexLocked = false; sMappedCount--; } @@ -1806,16 +1934,16 @@ void LLVertexBuffer::unmapBuffer() glUnmapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB); stop_glerror(); - mMappedIndexData = NULL ; + mMappedIndexData = NULL; } - mIndexLocked = FALSE ; + mIndexLocked = false; sMappedCount--; } if(updated_all) { - mEmpty = FALSE; + mEmpty = false; } } @@ -1835,12 +1963,12 @@ template <class T,S32 type> struct VertexBufferStrider if (ptr == NULL) { llwarns << "mapIndexBuffer failed!" << llendl; - return FALSE; + return false; } strider = (T*)ptr; strider.setStride(0); - return TRUE; + return true; } else if (vbo.hasDataType(type)) { @@ -1851,18 +1979,18 @@ template <class T,S32 type> struct VertexBufferStrider if (ptr == NULL) { llwarns << "mapVertexBuffer failed!" << llendl; - return FALSE; + return false; } strider = (T*)ptr; strider.setStride(stride); - return TRUE; + return true; } else { llerrs << "VertexBufferStrider could not find valid vertex data." << llendl; } - return FALSE; + return false; } }; @@ -1961,7 +2089,7 @@ bool LLVertexBuffer::bindGLBuffer(bool force_bind) glBindBufferARB(GL_ARRAY_BUFFER_ARB, mGLBuffer); sGLRenderBuffer = mGLBuffer; sBindCount++; - sVBOActive = TRUE; + sVBOActive = true; if (mGLArray) { @@ -1993,7 +2121,7 @@ bool LLVertexBuffer::bindGLIndices(bool force_bind) sGLRenderIndices = mGLIndices; stop_glerror(); sBindCount++; - sIBOActive = TRUE; + sIBOActive = true; ret = true; } @@ -2015,7 +2143,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) LLMemType mt2(LLMemType::MTYPE_VERTEX_SET_BUFFER); //set up pointers if the data mask is different ... - BOOL setup = (sLastMask != data_mask); + bool setup = (sLastMask != data_mask); if (gDebugGL && data_mask != 0) { //make sure data requirements are fulfilled @@ -2049,21 +2177,17 @@ void LLVertexBuffer::setBuffer(U32 data_mask) if (mGLArray) { bindGLArray(); - setup = FALSE; //do NOT perform pointer setup if using VAO + setup = false; //do NOT perform pointer setup if using VAO } else { - if (bindGLBuffer()) - { - setup = TRUE; - } - if (bindGLIndices()) - { - setup = TRUE; - } + const bool bindBuffer = bindGLBuffer(); + const bool bindIndices = bindGLIndices(); + + setup = setup || bindBuffer || bindIndices; } - BOOL error = FALSE; + bool error = false; if (gDebugGL && !mGLArray) { GLint buff; @@ -2072,7 +2196,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Invalid GL vertex buffer bound: " << buff << std::endl; } else @@ -2088,7 +2212,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { if (gDebugSession) { - error = TRUE; + error = true; gFailLog << "Invalid GL index buffer bound: " << buff << std::endl; } else @@ -2110,7 +2234,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) #endif sGLRenderArray = 0; sGLRenderIndices = 0; - sIBOActive = FALSE; + sIBOActive = false; } if (mGLBuffer) @@ -2119,13 +2243,13 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0); sBindCount++; - sVBOActive = FALSE; - setup = TRUE; // ... or a VBO is deactivated + sVBOActive = false; + setup = true; // ... or a VBO is deactivated } if (sGLRenderBuffer != mGLBuffer) { sGLRenderBuffer = mGLBuffer; - setup = TRUE; // ... or a client memory pointer changed + setup = true; // ... or a client memory pointer changed } } if (mGLIndices) @@ -2134,7 +2258,7 @@ void LLVertexBuffer::setBuffer(U32 data_mask) { glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0); sBindCount++; - sIBOActive = FALSE; + sIBOActive = false; } sGLRenderIndices = mGLIndices; @@ -2222,25 +2346,28 @@ void LLVertexBuffer::setupVertexBuffer(U32 data_mask) { S32 loc = TYPE_WEIGHT; void* ptr = (void*)(base + mOffsets[TYPE_WEIGHT]); - glVertexAttribPointerARB(loc, 1, GL_FLOAT, FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT], ptr); + glVertexAttribPointerARB(loc, 1, GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT], ptr); } if (data_mask & MAP_WEIGHT4) { S32 loc = TYPE_WEIGHT4; void* ptr = (void*)(base+mOffsets[TYPE_WEIGHT4]); - glVertexAttribPointerARB(loc, 4, GL_FLOAT, FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT4], ptr); + glVertexAttribPointerARB(loc, 4, GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_WEIGHT4], ptr); } if (data_mask & MAP_CLOTHWEIGHT) { S32 loc = TYPE_CLOTHWEIGHT; void* ptr = (void*)(base + mOffsets[TYPE_CLOTHWEIGHT]); - glVertexAttribPointerARB(loc, 4, GL_FLOAT, TRUE, LLVertexBuffer::sTypeSize[TYPE_CLOTHWEIGHT], ptr); + glVertexAttribPointerARB(loc, 4, GL_FLOAT, GL_TRUE, LLVertexBuffer::sTypeSize[TYPE_CLOTHWEIGHT], ptr); } - if (data_mask & MAP_TEXTURE_INDEX) + if (data_mask & MAP_TEXTURE_INDEX && + (gGLManager.mGLSLVersionMajor >= 2 || gGLManager.mGLSLVersionMinor >= 30)) //indexed texture rendering requires GLSL 1.30 or later { +#if !LL_DARWIN S32 loc = TYPE_TEXTURE_INDEX; void *ptr = (void*) (base + mOffsets[TYPE_VERTEX] + 12); - glVertexAttribPointerARB(loc, 1, GL_FLOAT, GL_FALSE, LLVertexBuffer::sTypeSize[TYPE_VERTEX], ptr); + glVertexAttribIPointer(loc, 4, GL_UNSIGNED_BYTE, LLVertexBuffer::sTypeSize[TYPE_VERTEX], ptr); +#endif } if (data_mask & MAP_VERTEX) { diff --git a/indra/llrender/llvertexbuffer.h b/indra/llrender/llvertexbuffer.h index e1cbfd3b61..ceaa70c3b6 100644 --- a/indra/llrender/llvertexbuffer.h +++ b/indra/llrender/llvertexbuffer.h @@ -55,9 +55,15 @@ class LLVBOPool { public: static U32 sBytesPooled; + static U32 sIndexBytesPooled; - U32 mUsage; - U32 mType; + LLVBOPool(U32 vboUsage, U32 vboType) + : mUsage(vboUsage) + , mType(vboType) + {} + + const U32 mUsage; + const U32 mType; //size MUST be a power of 2 volatile U8* allocate(U32& name, U32 size); @@ -88,7 +94,7 @@ public: //============================================================================ // base class -class LLPrivateMemoryPool ; +class LLPrivateMemoryPool; class LLVertexBuffer : public LLRefCount { public: @@ -103,6 +109,7 @@ public: }; LLVertexBuffer(const LLVertexBuffer& rhs) + : mUsage(rhs.mUsage) { *this = rhs; } @@ -118,9 +125,9 @@ public: static LLVBOPool sStreamIBOPool; static LLVBOPool sDynamicIBOPool; - static BOOL sUseStreamDraw; - static BOOL sUseVAO; - static BOOL sPreferStreamDraw; + static bool sUseStreamDraw; + static bool sUseVAO; + static bool sPreferStreamDraw; static void initClass(bool use_vbo, bool no_vbo_mapping); static void cleanupClass(); @@ -201,7 +208,7 @@ protected: void destroyGLIndices(); void updateNumVerts(S32 nverts); void updateNumIndices(S32 nindices); - virtual BOOL useVBOs() const; + bool useVBOs() const; void unmapBuffer(); public: @@ -239,8 +246,8 @@ public: bool getClothWeightStrider(LLStrider<LLVector4>& strider, S32 index=0, S32 count = -1, bool map_range = false); - BOOL isEmpty() const { return mEmpty; } - BOOL isLocked() const { return mVertexLocked || mIndexLocked; } + bool isEmpty() const { return mEmpty; } + bool isLocked() const { return mVertexLocked || mIndexLocked; } S32 getNumVerts() const { return mNumVerts; } S32 getNumIndices() const { return mNumIndices; } @@ -254,7 +261,7 @@ public: volatile U8* getMappedIndices() const { return mMappedIndexData; } S32 getOffset(S32 type) const { return mOffsets[type]; } S32 getUsage() const { return mUsage; } - BOOL isWriteable() const { return (mMappable || mUsage == GL_STREAM_DRAW_ARB) ? TRUE : FALSE; } + bool isWriteable() const { return (mMappable || mUsage == GL_STREAM_DRAW_ARB) ? true : false; } void draw(U32 mode, U32 count, U32 indices_offset) const; void drawArrays(U32 mode, U32 offset, U32 count) const; @@ -274,18 +281,25 @@ protected: S32 mSize; S32 mIndicesSize; U32 mTypeMask; - S32 mUsage; // GL usage + + const S32 mUsage; // GL usage + U32 mGLBuffer; // GL VBO handle U32 mGLIndices; // GL IBO handle U32 mGLArray; // GL VAO handle volatile U8* mMappedData; // pointer to currently mapped data (NULL if unmapped) volatile U8* mMappedIndexData; // pointer to currently mapped indices (NULL if unmapped) - BOOL mVertexLocked; // if TRUE, vertex buffer is being or has been written to in client memory - BOOL mIndexLocked; // if TRUE, index buffer is being or has been written to in client memory - BOOL mFinal; // if TRUE, buffer can not be mapped again - BOOL mEmpty; // if TRUE, client buffer is empty (or NULL). Old values have been discarded. - mutable BOOL mMappable; // if TRUE, use memory mapping to upload data (otherwise doublebuffer and use glBufferSubData) + + U32 mMappedDataUsingVBOs : 1; + U32 mMappedIndexDataUsingVBOs : 1; + U32 mVertexLocked : 1; // if true, vertex buffer is being or has been written to in client memory + U32 mIndexLocked : 1; // if true, index buffer is being or has been written to in client memory + U32 mFinal : 1; // if true, buffer can not be mapped again + U32 mEmpty : 1; // if true, client buffer is empty (or NULL). Old values have been discarded. + + mutable bool mMappable; // if true, use memory mapping to upload data (otherwise doublebuffer and use glBufferSubData) + S32 mOffsets[TYPE_MAX]; std::vector<MappedRegion> mMappedVertexRegions; @@ -296,28 +310,32 @@ protected: void placeFence() const; void waitFence() const; + static S32 determineUsage(S32 usage); private: - static LLPrivateMemoryPool* sPrivatePoolp ; + static LLPrivateMemoryPool* sPrivatePoolp; public: static S32 sCount; static S32 sGLCount; static S32 sMappedCount; - static BOOL sMapped; + static bool sMapped; typedef std::list<LLVertexBuffer*> buffer_list_t; - static BOOL sDisableVBOMapping; //disable glMapBufferARB - static BOOL sEnableVBOs; + static bool sDisableVBOMapping; //disable glMapBufferARB + static bool sEnableVBOs; static S32 sTypeSize[TYPE_MAX]; static U32 sGLMode[LLRender::NUM_MODES]; static U32 sGLRenderBuffer; static U32 sGLRenderArray; static U32 sGLRenderIndices; - static BOOL sVBOActive; - static BOOL sIBOActive; + static bool sVBOActive; + static bool sIBOActive; static U32 sLastMask; static U32 sAllocatedBytes; + static U32 sAllocatedIndexBytes; + static U32 sVertexCount; + static U32 sIndexCount; static U32 sBindCount; static U32 sSetCount; }; |